RAMAN AND PHOTOLUMINESCENCE STUDIES O N THERMALLY
ANNEALED POROUS SILICON
Ariushree lto.y+, I<. Jayararn and A. K. Sood"
Depiirtment of Physics, Indian Institute of Science,
Bangalore 560 012, India.
+Also at Solid State and structural Chemistry Unit,
Indiac Institute of Science, Bangalore 560 012, India.
"Also at Jawaharlal h'ehru Centre for Advanced Scientific Research.
11% Campus, Bangalore 560 012, India.
We report Rainan and Photoliiinincscciice (PI,) st udics of porous silicoit (PS) as a.
function of isochronal thermal ariiiealiiig from ruuin teiri~~er.atnre
to Y0O"C:. 'l'lic PI,
peak position atid intensity show Iiuriinotiotoiiic- variiitioii with incrcming terriperature. The PL intensity first increases upto 100°C, tlicn decreases till 550°C: and
recovers its intensity at 700°C before it corripletely disappears at 800°C. 'I'hta red
shifted asymmetric Ranian line shape can be fitted by phonon confinement model
along with the disordered silicoti componeitt. O u r rcsiilLs clearly indicate thitl. the
origin of visible PL can be better explained by il iiew hybrid rnodel which inrorporates both nariostructures for quantuiii cmifiiicrnetit arid silicoii ~:oniplc:xes(sucli as
SiH, and siloxerie) aiid d e f d s at Si/Si02 iiit(~II'iir~.~s
as I i i i i i i i i ( ~ ~ : t : i i ceiitrcs.
t
1. Introduction
The recent observations of highly efficient visible photo- and electro- luminescence at room temperature from
electrochemically etched porous silicon (PS) have stimulated a lot of excitement mainly due to the possibility
of its use in Si-based optoelectronic applications. But
till date, our understanding of the origin of the photoluminescence (PL) in PS is still incomplete. In one
model, the quantum confinement effect on the silicon
band structure arising from the formation of nanoscale
units ( < 50 h;) in PS is responsible for the visible PL.
The alternative model proposes that the PL is extrinsic to silicon and originats from silicon complexes like
siloxene [ 11, hydride species SiH, [2], polysilanes [3], or
hydrogenated amorphous silicon [4].
The quantum confinement model, first suggested by
Canham [5],is supported by transmission electron microscopy (TEM) studies and first principle electronic
structure calculations for H-terminated Si- wires (width
12 to 35 h;) [6]. The latter have shown that the band
gap of the silicon wire increases and become direct at
the zone centre. EPR measurements also supported the
confinement model [7]. However, there are many experimental observations involving variations of the PL
characteristics as a function of temperature, pressure,
laser irradiation and post-preparation treatments which
can not be explained in terms of the quantum confinement model. These are briefly mentioned as follows: (a)
The temperature dependence of the PL band in PS is
anomalous [8,9]. The peak position shows red as well as
blue shift with increasing temperature. Rapid thermal
oxidation experiments [lo] have shown red shift of the
peak till 650°C and then blue shift for 650°C < T <
900°C . The PL integrated intensity decreases between
200°C and 3OO0C,and then recovers at 750°C. (b) Prokes
el a1 [ll]have carried out experiments on thermal annealing of PS at 430°C followed by dipping it in HF
acid. During the heating cycle, the PL intensity drops
significantly and the peak shifts to 1.68 eV. The HF dipping restores the intensity and the peak value shifts to
1.90 eV. (c) Various groups have reported the contradicting results on the effect of the hydrostatic pressure
on the peak position of the PL. Sood et a1 [12] have
reported red shift of the peak position with increasing
pressure whereas Zhou et a1 [13] observe a blue shift for
low pressure and then a small red shift above 20 kbar.
(d) Laser irradition causes a blue or a red shift of the
PL peak position depending on the parameters of the
sample preparation [13,14]. (e) Increasing porosity did
not always correspond to a blue shift of the PL band
[15]. (f) The expected correlation between red shift of
the Raman line due to phonon confinement and the corresponding blue shift of the PL band in nanostructure is
shown to be absent [12]. (g) Photoemission studies have
shown presence of Si, 0 and F in ion-etched PS [IS].
Keeping in mind the incomplete, rather confused,
understanding of the origin of the PL in PS, we have
carried out simultaneous Raman and PL studies on PS
as a function of isochronal thermal annealing from room
temperature to 900°C. Our other motivation to carry out
thermal annealing studies came from the recent report of
electroluminescence in diffused PS p-n junction, where
the diffusion of phosphorous is carried out at 850°C for
30 minutes [17]. The earlier Raman measurements on
PS [12] have shown that the phonon is red shifted with
respect to the bulk silicon and the line shape is highly
asymmetric on the low frequency scale. This has been
attributed to confinement of the phonon in nanoscale silicon particles formed in the PS. Our objective of Raman
measurement is to see the effect of thermal annealing on
the nanoscale structures together with changes in luminescence characteristics. Our results support a new
hybrid model [14] which incorporates features of both
the models of quantum confinement as well as the silicon based compounds like SiH, and SiOz formation in
the PS.
Soot
2. Experimental
The porous silicon was prepared by anodic etching
of boron doped p-type Si wafers of < 100 > orientation
(resistivity 1 R-cm) in 48% HF : ethanol (1:l) solution,
keeping the current density between 10 and 15 mA/cm2.
PL and Raman spectra were recorded at room temperature in 45" reflection geometry using 5145 8, line of
argon ion laser as excitation source and computer controlled SPEX Ramalog (model 14018) equipped with a
cooled photomultiplier tube and a photon counting system. For thermal treatment in air, the sample was introduced in the temperature controlled furnace maintained
at a desired annealing temperature (T,) and kept for
5 minutes. Then it was taken out of the furnace and
cooled to room temperature to record both Raman and
PL spectra at the same spot of the sample. The sample
was annealed at seventeen different temperatures ranging from 25°C to 900°C with temperature interval of
50°C. The experiments were repeated on three samples
to check the reproducibility of the reported results.
3. Results
In Fig. 1 we show the PL of PS at nine different
temperatures. The intensity scale in the figure is same
for all spectra and hence one can clearly see the nonmonotonic variation of intensity of PL with temperatures. The measured PL spectra can be very well fitted
by Gaussian profile as can be seen in the lowest spectrum of Fig. 1, corresponding to PL of as-prepared PS
(T, = 25°C). The peak positions and intensities (area
under the band) of PL as obtained from the non-linear
least square fitting of the data with the Gaussian profile
are shown in Fig. 2 as a function of annealing temperature T,. The intensities are normalized with respect
to that of the as-prepared PS. The solid lines in Fig. 2
have been drawn through the points as guide to the eye.
It can be seen from Fig. 2(a) that the PL peak energy
is 1.73 eV in as-prepared PS and increases to 1.77 eV at
T, = 100°C and then gradually red-shifts to 1.68 eV at
T, = 600°C. Above 600°C the peak energy increases and
recovers to 1.74 eV (T, = 750°C). In Fig. 2(b) we see
.-
"
JIE
15
1.6
1.7
1.8
19
2.0
2.1
22
Fig. 1. Photoluminescence spectra of porous silicon at
nine different annealing temperatures. Intensity scale of
all spectra is same. Horizontal line on the right of each
spectrum shows the zero of the intensity scale. In the
lowest curve the dots represent experimental data and
the solid line is the fitted Gaussian lineshape.
that the PL intensity remains constant till T, reaches
100°C and then it increases upto ?', = 350°C, followed
by a decrease. Above 550°C the intensity of PL from
PS again starts increasing and then decreases for T, >
700°C. At 800°C, the intensity vanishes completely.
The corresponding changes in Raman spectra as a
function of annealing temperature are shown in Fig. 3.
In the same figure we have also shown (top curve) the
Raman spectrum of bulk crystalline Si (c-Si). Comparing the Raman spectra from PS at different T, with
that of c-Si, one can see that Raman line of the PS is
red-shifted and highly asymmetric till T, = 550°C. It is
known that the asymmetric Raman line shape can arise
due to the phonon confinement in the nanoscale crystallites [18]. In PS these can be either approximated to
cylindrical (ie. quantum wire) or spherical in shape. In
this model the Raman line shape is given by
where w ( a ) is the phonon dispersion curve for bulk
silicon, which has been taken as w(p) = wo(l - 0.18q2),
which fits the experimental curve well in the direction
of r - X upto p = 0.8. w, and rc are phonon frequency
= 0) and the natural linewidth of the bulk
(a
--- Phonon confinement lineshepe I,
- Combinedlineshapel
n
0
Fig. 2. The variation of the phtoluminescence (a) peak
position and (b) intensity of porous silicon as a function
of annealing temperature. The intensities are normalized with respect to that of as-prepared porous silicon.
The solid lines through the data points are guide to the
eye.
Fig. 4. Dots show the Raman spectrum of as-prepared
porous silicon. The dashed line shows the fit using the
confinement model alone Ic(w). The solid line shows the
best fit using I ( w ) = I c ( w ) +Id(w) (a sum of phonon
confinement effect and disorderd component).
,4
"
f
Fig. 3. Raman spectra of porous silicon at eight different annealing temperatures. The Raman line of the
bulk crystalline Si is shown at the top.
Fig. 5. Variation of the (a) the relative intensity of
two components of Raman line, B / A (b) Raman peak
frequency of the disordered component, wd (c) Raman
linewidth (HWHM) of the disordered component , r d
(d) nanostructure diameter, L with different annealing
temperatures. The solid lines through the points are
guide to the eye.
silicon, the values of which have been taken as w, =
521.9 cm-' and r, = 4.7 cm-'. The wavevector q has
been expressed in units of 27r/a (where a is the lattice
constant of Si = 5.4A) and qmDI = 0.8. A is a constant.
c(0,q') is the Fourier coefficient of the phonon confinement function. One may note that the Raman line in
nanoparticles is red-shifted and asymmetric towards the
low frequency side because the phonon dispersion w( 4
in bulk silicon is a decreasing function of q . We have
used the phonon confinement function to be Gaussian
which is found to be suitable for micro-crystalline Si :
I c(0,q' ) 1' = e z p [-qZLz/(4a2)],L is the dimension
of the nanocrsytallite of Si. L is the diameter of the
cylinder for infinitely long cylindrical particles and diameter for spherical particles. It has been shown that
for the PS, the phonon confinement in spherical particles explains the line shape better than with cylindrical
particles [19,12]. Hence we have tried to fit the experimental Raman spectra from the PS using the above
model with spherical particles. The best fit with diameter L = 42.7A is shown by dashed line in Fig. 4 for
the as-prepared PS and the corresponding value of x2 is
2304. It is obvious from Fig. 4 that the phonon confinement model alone can not explain the PS Raman line
shape.
In the next step, we attempted to fit the observed
Id(w), where z d ( W ) =
line shape with I ( w ) = zc(w)
Brd[(w - wd)'+ r;]-' is the Lorenzian line shape and
arises due to disordered/amorphous component of the
Ps. Here wd and r d are the phonon frequency and half
width at half maximum (HWHM), respectivly for the
disordered component and B is a constant. The presence
of the disordered/amorphous phase has been reported
earlier using EPR measurements [ll]. Raman scattering measurements of microcrystalline silicon prepared
in hydrogen plasma have shown crystalline as well as
amorphous features. The latter has been attributed to
a surface-like Si - Si shearing mode at the grain boundaries of the microcrystallites [20]. In the above exercise
of nonlinear least square fit of the data with I ( w ) , the
fitting parameters are L , wd, r d A and B . In Fig. 4
we have plotted the fitted I(w)(solid line) and the corresponding value of xz is 350. The value of L is the
same (42.7 8) as obtained by the previous fit with Zc(w)
alone. It can be seen from Fig. 4 and the lower value of
xz that the Raman line shape Z(w) = Zc(w) Zd(w) fits
extremely well the experimental spectrum. In Fig. 5(a)
we show the relative intensity B / A of the two components I d ( W ) and Zc(w). In Fig. 5(b-d) we have shown the
variation of wd, r d and L with different T,. The solid
lines through the points are guide to the eye.
+
+
4. Discussion
Our experimental results can not be explained in
terms of either of the two models mentioned earlier. Instead a new hybrid model [14] may be more suitable to
understand the experimental data. According to this
model the incident radiation creates electron-hole pairs
in nanoscale structures present in the PS but these carriers, unlike in quantum confinement model, do not radiatively recombine inside the nanoparticles. Instead,
these carriers diffuse out and recombine through various luminescent centres. These centres can be point defects located at the Si/SiOz interface or in the thin SiOz
layers or complexes of silicon with 0, H or F formed
on the surface of the nanoparticles. The dominance of
any of these luminescence centres will depend on the
anode-etching conditions, post-preparation and thermal
treatments. This can rationalize the different reports
on variations of the PL characteristics in the PS. In our
experiments, initial increase in the PL intensity with
annealing temperature in Fig. 2(b) suggests an increase
in defect density during thermal annealing. As PS is
heated beyond 350"C, hydrogen desorbs from the sample
as seen in EPR studies (via increase in dangling bonds)
and infra-red absorption measurements [ll]. This will
decrease the defects related to Si-H complexes at the
surface of the nanostructures and hence the PL intensity. This can also cause a decrease in the PL peak
position. Further heating (T, > 500°C) results in the
repassivation of the dangling bonds due to formation of
SiOz. The luminescence now can be due to the defects
at Si/SiOz interface or the defects in SiOz and can result
in higher PL peak position as seen in Fig. 2(a). Now,
coming to Raman scattering results, we note the following : Variation of the relative intensity ( B / A )[ intensity
of Zd(w)/intensity of I, (w)] in Fig. 5(a) agrees with the
above conclusion that the disorder in the form of possible Si-H complexes first increases upto 350°C and then
decreases. The variation in wd and r d in Fig. 5(b) and
(c) with annealing temperature suggests that composition and structure of the Si-H complexes are temperature sensitive coroborating the corresponding changes in
the PL peak position. The drastic change in the Raman
line shape for T, > 550°C can arise because increasing
portions of nanoscale Si get converted into SiOz. The
fact that the SiOz has much less absorption coeffecient
of the incident radition at 2.41 eV (as observed in the
shift of the absorption band at T,- 600°C [21]) the Raman signal will be dominated by the underlying substrate. Fig. 5(d) suggests that the size of the nanoscale
unit changes with annealing temperature. The initial
increase upto 100°C seems t o be correlated with the increase of wd [Fig. 5(b)] from 487 to 502 cm-', similar to
the reported dependence of the peak of the amorphouslike Raman component on the size of the microcrystallite
of Si [20]. More quantitative TEM studies are required
to look into this aspect.
In conclusion, the characteristics of both the nanostructures and the defect states responsible for PL in the
PS have been investigated as a function of thermal annealing. PL experiments help in understanding the nature of the luminescent states whereas Raman scattering probes the underlying quantum network. A hybrid
model incorporating the features of nanoscale units acting as a reservoir of the electron-hole pairs and silicon
complexes (with 0, H and F) and defects located outside the nanoparticles as radiative luminescent centres
can explain our experimental results reasonably well.
Acknowledgement - One of us (AKS) thanks Department
of Science and Technology for financial support.
References
[l] M. S. Brandt, €1. D. Fuchs, M. Stutzmann, J. Weber and M. Cardona, Solid State Commun. 81, 307
(1992).
[2] L. E. Friedersdorf, P. C. Searson, S. M. Prokes, 0.
J. Glembocki and J. M. Macaulay, Appl. Phys. Lett.
60,2285 (1992).
[3] M. B. Robinson, A. C. Dillon, D. R. Haynes and S.
M. George, Appl. Phys. Lett. 61,1414 (1992).
141 D. J. Wolford, B. A. Scott, J. A. Reimeer and J. A.
Bradley, Physica 117B and 118B,9220 (1983).
[5] L. T. Canham, AppZ. Phys. Lett. 57,1046 (1990).
[6] A. J. Read, R. J . Needs, K. J. Nash, L. T. Canham,
P. D. J. Calcott and A. Qteish, Phys. Rev. Letts.
69,1232 (1992).
[7] S. V. Bhat, K. Jayaram, D. V. S. Muthu and A. I<.
Sood, Appl. Phys. Lett, 60,2116 (1992).
[8] X. L. Zheng, W. Wang and H. C. Chen, Appl. Phys.
Lett. 60,986 (1992).
[9] Z. Y. Xu, M. Gal and M. Gross, Appl. Phys. Lett.
60,1375 (1992).
[lo] K. H. Li, C. Tsai, J. C. Campbell, B. K. Hance and
J. M. White, Appl. Phys. Lett. 62,3501 (1993).
[ll] S. M. Prokes, W. E. Carlos and V. M. Bermudez,
Appl. Phys. Lett. 61,1447 (1992).
[I21 A. K. Sood, K. Jayaram and D. Victor S. Muthu,
J. Appl. Phys. 72,4963 (1992).
[13] W. Zhou, H. Shen, J. F. Harvey, R. A. Lux, M.
Dutta, F. lu, C. H. Perry, R. Tsu, N. M. Kalkhoran
and F. Namavar, Appl. Phys. Lett. 61,1435 (1992).
[14] G. G. Qin and Y. Q. Jia, Solid State Commun., 86,
559 (1993).
[15] S. M. Prokes, 0. J. Glembocki, V. M. Bermudez,
R. Kaplan, L. E. Friedersdorf and P. C. Searson,
Phys. Rev. B 45 13788 (1992).
[16] A. Roy, A. Chainani, D.D.Sarma and A.K. Sood,
Appl. Phys. Lett. 61,1665 (1992).
[17] V. K. Jain, Amita Gupta, A. Kumar, G. K. Singhal,
0. P. Arora, T.Srinivasan, D. S. Ahuja, P. Puri, and
Vikram Kumar, BUZZ.Mater. Sci. 16,239 (1993).
(181 I. H. Campbell and P. M. Fauchet, Solid State Commun. 5 8 , 739 (1986).
[19] Z. Sui, P. P. Leong, I. P. Herman, G. S. Higashi and
H. Temkin, Appl. Phys. Lett. 60,2086 (1992).
[20] Z. Iqbal and S. Vepfek, J. Phys. C : Solid State
Phys. 15, 377 (1982).
[21] V. Petrova-Koch, T. Muschik, A. Kux, B. K.
Meyer, F. Koch and V. Lehmenn, Appl. Phys. Lett.
61,943 (1992).